M. Colaianni et al., "A Pose Invariant Statistical Shape Model for Human Bodies", in Proc. of 5th Int. Conf. on 3D Body Scanning Technologies, Lugano, Switzerland, 2014, pp. 327-336, doi:10.15221/14.327.
A Pose Invariant Statistical Shape Model for Human Bodies
Matteo Colaianni 1, Michael Zollhöfer 1, Jochen Süssmuth 2, Bettina Seider 2, Günther Greiner 1
1 Computer Graphics Group, University Erlangen-Nurnberg, Germany
2 Adidas Group, Germany
We present a complete pipeline for constructing a statistical shape model that is invariant to deviations in the scan pose while encoding the space of human pose and body shape in an efficient manner. A dense cross-parameterization between a large set of high-quality 3D scans is computed using a fast and robust volume aware non-rigid registration method. Our approach uses a novel encoding that automatically decorrelates shape and pose leading to a statistical model that is oblivious under transformations induced by pose. This allows us to efficiently compensate pose variations in captured input data leading to a compact representation for pose as well as body shape. We present a local as well as a global skeletal encoding and compare both approaches. Finally, we analyze the generalization properties and accuracy of our approach against two state-of-the-art methods. We apply our model to the data clustering problem and use it as a prior for non-rigid shape matching.
© Hometrica Consulting - Dr. Nicola D'Apuzzo, Switzerland, www.hometrica.ch.
Reproduction of the proceedings or any parts thereof (excluding short quotations for the use in the preparation of reviews and technical and scientific papers) may be made only after obtaining the specific approval of the publisher. The papers appearing in the proceedings reflect the author's opinions. Their inclusion in these publications does not necessary constitute endorsement by the editor or by the publisher. Authors retain all rights to individual papers.